The midsize firms are moving and you're not. A new Epicor survey of over 1,000 frontline workers found that companies with 50 to 999 employees are adopting AI faster than both large enterprises and very small firms. That's the uncomfortable middle you might be below. The reason makes sense when you think about it: midsize companies have enough money to invest, enough complexity for AI to actually solve real problems, and small enough teams to skip the six-month approval process. The firms least likely to have any AI plans at all? Those with 1 to 49 employees, and those with 5,000 or more. You're in the same bucket as the slow giants. That's not great. Read the full breakdown here.
Before you buy an AI agent, read this. "Agentic AI" is the pitch you're hearing everywhere right now. An AI agent is basically software that doesn't just answer questions but actually goes and does things on your behalf, like a very diligent assistant who never sleeps. The promise is real. But a new overview from MSN gives a useful reality check: if your data is messy, your processes aren't documented, and your team doesn't understand what the agent is doing, the thing will fail. Not maybe fail. Over 40% of agentic AI projects are expected to fail by 2027, according to the piece. The checklist before you spend anything: clean data, written-down processes, and a team that understands what they're handing off. If those aren't in place, you're not automating your business, you're automating your chaos. Full readiness checklist here.
Tools Worth a Look
Plaud One: Wearable device with a built-in SIM that captures conversations and turns them into meeting notes and action items, hands-free. For owners always on the move. $249.99 pre-order (ships Q4 2026, includes $200 in Plaud credits).
Meta Muse Image API: Generate or edit images via API at $0.01 per image. For owners or developers who want to embed image creation into their own tools without paying premium prices.
GLM-5.3-Flash by Z AI: A capable open-weight AI model at roughly $0.045 per task, about 10x cheaper than comparably ranked models. For technical teams building AI-powered workflows on a budget.
ChatGPT Scheduled Tasks (trigger update): ChatGPT scheduled tasks can now fire from external events like a new email or Slack message, not just a set clock time. Included with Pro, Plus, and Business plans.
DigitalOcean M.A.R.S.: Managed cloud environment for running AI coding agents (Claude Code, Codex, and others) in isolated sandboxes with built-in access to GitHub, Jira, Notion, and Postgres. For small dev teams who want to run agents without managing the infrastructure themselves.
Trending on GitHub
bilawalsidhu/gods-eye-view: A live, photorealistic 3D globe in your browser that pulls real open-source satellite and spatial data. Think of it as a spy satellite simulator running on public information. (1,984 stars today)
DietrichGebert/ponytail: A tool that steers AI coding agents toward the simplest possible solution, less code written, fewer problems to maintain. (1,613 stars today)
K-Dense-AI/scientific-agent-skills: A library of 163 ready-to-use skills that plug into AI agents for science-related tasks, covering biology, chemistry, medicine, and drug discovery. Compatible with Cursor, Claude Code, and Codex. Used by over 175,000 scientists. (498 stars today)
JetBrains/go-modern-guidelines: A style guide published by JetBrains to help AI coding agents write current, idiomatic Go code instead of outdated patterns. (300 stars today)
zedeus/nitter: An alternative way to browse Twitter/X without the native interface, tracking, or login walls. (71 stars today)
More Headlines
LegalZoom survey: entrepreneurs use AI for speed on first drafts, but hand off to humans when real liability is on the line. Useful reminder that AI handles the prep work, not the final call on contracts or disputes.
Meta's autonomous AI agent "Hatch" could launch within weeks, priced as high as $199.99/month. If the pricing holds, it's aimed at owners who want a capable agent, not just a chatbot.
Apple's September 9 event will debut an LLM-powered Siri on 2.5 billion active devices. If Siri becomes the layer customers use for search, bookings, and purchases, your business needs to think about how it shows up to an AI assistant, not just a search engine.
OpenAI's in-house "Jalapeño" chip outperforms Nvidia Blackwell by 3.6x on speed and 1.9x on efficiency per watt, with API price cuts already flowing through. Cheaper inference means lower costs for anyone building on OpenAI's API.
How AI-driven cash flow forecasting and reporting is changing the relationship between small businesses and their banks. Worth knowing, especially if you're a client wondering why your bank's interface suddenly looks different.
-Scott
SmartOwner is published (almost) daily by the team at DigitalTreehouse. Want AI consulting or automations for your business? Reply to this email.